AutoCode Finds 600% More Grade 4 & 5 Defects Than Manual Coding in Village of Schaumburg Accuracy Study
In a head-to-head comparison of ~75,000 linear feet of inspection data, SewerAI AutoCode identified more than 600% more Grade 4 & 5 defects per mile and over 1,400% more structural defects per mile than the Village of Schaumburg's self-performed manual PACP coding — validated alongside RJN Group, the Village's engineer of record.
Overview
The Village of Schaumburg, Illinois partnered with SewerAI and RJN Group — the Village's engineer of record — to conduct a rigorous accuracy study comparing AI-assisted defect coding against the Village's own self-performed manual PACP coding. Across approximately 75,000 linear feet of inspection data processed through AutoCode, the results were unambiguous: manual coding was missing the vast majority of significant defects in the collection system.
The Study
The study processed ~75,000 linear feet of existing inspection footage through SewerAI's AutoCode platform and compared the results directly against the Village's self-performed manual PACP coding of the same footage. The analysis was conducted alongside RJN Group, providing independent engineering validation of the findings.
The core metric was "defects per mile" — a normalized measure that allows direct comparison of defect detection rates regardless of footage length. The analysis focused on two critical categories:
- Grade 4 & 5 defects (O&M and Structural) — the most severe defects requiring priority attention
- Structural defects per mile — conditions indicating pipe integrity risk
Results
Grade 4 & 5 Defects Per Mile: 3.24 vs. 22.89
Manual coding identified 3.24 Grade 4 & 5 defects per mile. AutoCode identified 22.89 — more than 600% more. These are the defects that demand the most urgent operational and capital response. Missing them means deferred repairs, increased risk of failure, and capital planning built on incomplete data.
Structural Defects Per Mile: 6.48 vs. 99.93
The gap was even more pronounced for structural defects specifically. Manual coding found 6.48 structural defects per mile. AutoCode found 99.93 — more than 1,400% more. Structural defects represent conditions that threaten the physical integrity of the pipe. Underreporting them at this scale means the Village's collection system contains far more structural risk than its manual inspection program had revealed.
What This Means
The Schaumburg accuracy study illustrates a challenge that is not unique to one municipality: manual PACP coding, even when performed by trained operators, systematically underreports defects — particularly the most severe ones. The consequences compound over time:
- Capital improvement programs are sized and prioritized on incomplete condition data
- Grade 4 & 5 defects go unaddressed until they become emergencies
- Structural risk accumulates invisibly across the collection system
By running the same footage through AutoCode, the Village of Schaumburg and RJN Group were able to see their collection system as it actually is — not as manual coding had represented it. That visibility is the foundation of effective infrastructure management.